IAmNotHere24/arbiter-gen

add SFT training plots

fix: shorten HF space description

chore: remove large PDF from tracking

chore: exclude results/plots and jsonl from Space deployment

chore: track binary/plot files via LFS

chore: exclude lora checkpoints from Space deployment

chore: track large tokenizer.json via LFS

done ig

changes

improved

feat: loan-to-any-domain refactor - DomainConfig v2, graph.py causal chain fix, demo Unicode fix, grpo_trainer import-safe

merge: vraj-feat into master — L6/L7 dual-auditor, schema drift, biased SFT, no data lost

fix: merge server.py - combine OpenEnv integration + L7 dual-auditor endpoints

rubrics changed

fix: handle TRL versions without DataCollatorForCompletionOnlyLM

grpo trainer fixed

feat: L6 schema drift + L7 dual auditor + biased SFT generator

feat: proper OpenEnv compliance — ArbiterEnvironment subclasses openenv.core.Environment

yay

hello

Merge pull request #3 from VrajVashi/bugfix

Merge branch 'master' into bugfix

Bugfix

Merge pull request #2 from VrajVashi/utk-features

Add React frontend, Dockerfile, OpenEnv integration tests and tasks

Merge pull request #1 from VrajVashi/kabir-feat

fix: SFT trajectory normalization and budget deduction bug

feat: add Groq + Gemini provider support to sft_generator

chore: update TASK_LIST.md to 65% completion + fix gitignore

feat: GRPO trainer, SFT script, 3-condition evaluator, visualization, FastAPI server, demo plots

feat: add graphify knowledge graph (51 nodes, 58 edges, 14 communities)

feat: add arbiter/env core modules

feat: complete ARBITER environment (70/70 validation pass)